Huawei engineers presented XMFS, an experimental Linux kernel prototype filesystem, at the Linux Plumbers Conference in Prague on October 5. It aims to let applications reach cross-node shared memory on CXL 3.0 or Huawei unified bus servers through standard POSIX file calls. The code exists only on openEuler, not in the mainline Linux kernel.
Perplexity has released pplx-embed-v2-late, a pair of ColBERT-style multimodal embedding models in 0.6B and 9B sizes that retrieve text, images and rendered PDF pages in a shared embedding space. Both are available on Hugging Face under the MIT license, while a hosted API endpoint is planned but not yet live.
Claude now turns company data into dashboards that stay current, and it can build animated explainers from a prompt. Dashboards connect to BigQuery, Databricks, Snowflake, and Salesforce in beta on paid plans, while Motion is in beta on Team and Enterprise. Docs, Slides, and Design are out of beta and available on every plan, including Free.
AIWhy it matters: The post specifies which data platforms connect, which features move out of beta, and where admins control access, clarifying what changes for enterprise workflows.
Johns Hopkins astrophysicist Brice Ménard, working as an Anthropic researcher, used Claude Science to produce the first complete map of the sky in ultraviolet light. Claude orchestrated agents to merge GALEX, Swift, and FIMS/SPEAR data, then predicted roughly a third of the sky that no UV telescope had observed, using relationships to visible, infrared, and radio data. Hidden test regions were reconstructed to within about 10% of real measurements, and each pixel is labeled measured or predicted with uncertainty estimates.
AIWhy it matters: The post shows how an astrophysicist used Claude Science agents to merge UV surveys and predict missing sky regions, with a validation step that makes the method reusable.
Memory has overtaken compute as the central control point in IT, according to The Next Platform, as generative and agentic AI drive demand for DRAM, HBM, and flash. Server DDR5 memory now sells for roughly 9X to 13X its November 2022 street price, while a 30 TB enterprise SSD costs 6X to 7X more. HBM pricing has risen only about 1.6X since the GenAI boom began, the article says.
François Chollet compares AI progress to human learning, noting that 1980s research found programming training improves coding but does not transfer to general reasoning. He argues general intelligence is a fundamental brain property rather than a trainable skill, since domain practice improves only that domain. The post is framed as background for his question whether AI's jagged frontier, driven by math and code via RLVR, reflects general capability or continued human-data bottlenecks.
Apple researchers introduced Normalizing Trajectory Models (NTM), which model each reverse diffusion step as a conditional normalizing flow trained with exact likelihood. The model matches or outperforms strong image generation baselines on text-to-image benchmarks in just four sampling steps while retaining exact likelihood over the generative trajectory.
Epoch gave six models 11 real work tasks from its own operations, including graphic design, data insights, and research design, and graded outputs against employee standards. Fable 5.1 and GPT-6 Astra led on average task performance, reliably handling well-defined work such as coding and computational analysis. The report finds that all models still fail on open-ended judgment, including matching Epoch's standards, designing informative experiments, and generating diverse ideas, so the authors conclude AI cannot yet replace workers at Epoch.
AIWhy it matters: The report separates well-defined task reliability from open-ended judgment failures, which benchmark scores on easily verifiable tasks would miss.
A Hugging Face blog author used the ML-Intern agent in HuggingChat to build six small models by writing detailed prompts that specify datasets, base models, baselines, smoke tests, and spending limits. The projects include a citrus disease vision-language model, a Huggy character LoRA, a camera-angle LoRA, a doodle-to-object LoRA, a 0.8B prompt rewriter, and a 4-step distilled Agate model, with total compute cost of about USD 103. Each project's prompts and public models are linked from the post.
AIWhy it matters: The author shows how prompt structure, baselines, smoke tests, and budget caps shape an agent-driven training workflow, with per-project costs given.
Databricks has put Workday Data Connect federation into Beta in Unity Catalog, letting teams query Workday HR and finance data without copying it. Workday Data Cloud customers get zero-copy, read-only access to the shared tables, with Databricks running queries and Unity Catalog governing access, lineage, and auditing. Teams can combine current people and financial data with other enterprise data for analytics and AI, including Genie-powered natural-language exploration.
If you want people to do open research on the impact of distillation to frontier progress, we happily would at @trillium_labs, but just need to scale up funding/compute before getting that done. Happy to make this happen sooner - you know where to find me!
Try invent: https://adaptionlabs.ai/invent-a-dataset Read this thread from @singhshiviii about the technical details: https://x.com/singhshiviii/status/2105711112880853414?s=20
Generated videos are getting so good. One of the @adaption_ai team put this together as an explainer of invent a dataset. 🔥 The biggest hurdle in frontier AI is curating high-quality, diverse data. Invent solves for this zero data regime.
Join Alex Bie at the @COLM_conf Google booth #107 today at 5:00 PM for a walkthrough of ContinuousBench, a standardized benchmark designed to measure knowledge transfer in differentially private (DP) synthesis. Don't miss the chance to explore if DP synthetic data truly preserve information, or just style? @GoogleDeepMind Read the paper: https://arxiv.org/abs/2606.01849
Today we announce new findings from Visiting Fellow David Autor on how AI impacts how professionals build expertise. In a three-month randomized controlled trial with practicing patent attorneys, we test both short-term productivity & longer-term skill building that occur as a result of AI usage. More: http://goo.gle/4AVeOWf
An AI agent makes a mistake early in a task, then keeps going in the wrong direction. Our researchers built PivotOPD to teach agents how to avoid those mistakes and recover when they happen. During training, a teacher model shows the agent a better action and how to get back on track over the next few steps. Read the paper and watch how it works: https://research.nvidia.com/labs/lpr/pivotopd
LLM agents learn by interacting with environments, but static setups limit their growth. Today at 2:00 PM, join Zifeng Wang at the #COLM2026 Google booth (#107) to learn about EnvHarness, a plug-in architecture that dynamically reshapes environment behaviors to improve reinforcement learning and agent adaptability.
Microsoft's Foundry blog guide advises keeping existing Azure Document Intelligence workloads that meet production requirements. It recommends evaluating Azure Content Understanding for high-variation, unstructured, reasoning, RAG, or multimodal document scenarios, and for new cloud OCR or layout workloads.
What if AI could figure out the right business context without waiting for a fully built ontology? Genie Ontology does that by inferring relevant context across your data and assets to answer open-ended questions. Advancing Analytics’ @MrSiWhiteley digs into how OntoRank chooses what to trust, why certified assets still matter, and what that means for data governance. Watch the full video: https://www.youtube.com/watch?v=voZXeHP2qWQ
🙌 Open models for open science!
NVIDIA's technical blog describes how its team taught robots to assemble GB300 tester trays, a task that currently requires skilled manual labor in factories. The post also discusses lessons about robot learning, mechanical intelligence, and engineering. The source excerpt provides limited detail beyond this.
We instructed the AI models that their technique should improve performance on several benchmarks. We already knew that all of these could be improved by a recent human-authored post-training innovation: on-policy self-distillation (SDPO).
We gave Fable 5 and GPT-5.6 Sol 3000 GPU-hours and tasked them with developing a novel post-training technique that would improve on a standard pre-existing baseline (GRPO).
AI developers aim to create an automated AI researcher. How close are they? To find out, we built InnovationEval, which tests whether AI can produce post-training innovations comparable in magnitude to a recently published advance. So far, agents’ results are underwhelming.
Another cool update from the robotics world! While we're all wow-ed in mental land, there's a ton of progress and i feel like acceleration happening in physical land too now. My guess is in good part also (but not only) thanks to the coding models progress over the last year.
GitHub reports that one in three pull requests now involves an AI agent, and that public secret exposures rise with the volume of pushes rather than from declining developer care. It introduces a ModernBERT-based classifier with Microsoft Applied Sciences that evaluates candidate secrets in under two milliseconds and could more than double the secrets prevented at push time. The feature is in private preview, with availability for GitHub Secret Protection customers later this month.
Worth reading — teach according to each student's aptitude! Sherpa: Teaching LLMs to Teach Adaptively
The post argues that companies are increasingly recognizing business opportunities from reinforcement learning, since many real-world tasks need specialized models, harnesses, and data flywheels rather than AGI. It predicts a new post-training era led by full-stack AI companies, though it provides no concrete figures or raw data to support the claim.
With just 2.5GB VRAM, you can train your own Decision Model using small models like Laya. Training is simply done through a UI interface. Video tutorial and analysis are in our guide. GitHub repo: https://github.com/unslothai/unsloth
Google launched its first test satellite carrying four TPUs into orbit last week as part of Project Suncatcher, a moonshot exploring whether machine learning infrastructure could one day operate in space. The test aims to determine whether Google's AI hardware can withstand the physical stress of spaceflight and the radiation and thermal extremes of orbit.
We made it possible to train your own Decision model locally on just 3GB VRAM! We converted Qwen, Gemma, Llama all into decision models by fine-tuning using a Clef head, boosting accuracy from 30 to up to 78%. You can try it yourself via Unsloth Desktop!
Unsloth released an open-source method to fine-tune LLMs into decision models that run locally, lifting Qwen3.5 0.8B's aggregate accuracy from 20.7% to 74.3% across three decision benchmarks. The team used a Clef head with LoRA (r=64) for one epoch on just 4GB VRAM, with the approach applicable to models such as Qwen3.8 and Gemma 4. A guide and notebooks are available on the Unsloth documentation site and GitHub.
This is very clever: It's a model that "extracts" values that don't exist in a document but can be derived from the values that do exist. For example, you can ask for the yearly cost of using a product, and it will automatically compute it from the monthly cost. The video shows a bunch of cool examples. It looks like it can infer really convoluted and sophisticated formulas.
Both models are on Hugging Face: https://huggingface.co/collections/perplexity-ai/pplx-embed-v2
Both sizes are distilled token by token from one 18B teacher, so they share one embedding space. A corpus indexed with the 9B model can be searched with 0.6B queries. That lifts ViDoRe v3 from 62.3% to 63.5% with no added query cost.
Dense embedding models compress each document into one vector, which loses detail as pages get longer or more visual. pplx-embed-v2-late keeps a 128-dimensional vector per token and scores with MaxSim, so each query token is matched to its closest token in the document.
LlamaIndex has launched OpenDocRouter, a unified API that transcribes documents to markdown across frontier and open-weight models at cost plus a small transaction fee. The service handles rate limits, adds bounding boxes and layout as a service, and will benchmark new OCR models on ParseBench before adding them.
LlamaIndex announced OpenDocRouter, a single API that routes document parsing requests to any of 10 frontier and open-source models at launch, including Claude Opus 5.5, Gemini 3.8 Flash, GPT-6 Luna, MinerU2.5-Pro, and PaddleOCR-VL-1.6. Users can switch models in one line with the same request and markdown output, and each model is scored on ParseBench for quality and cost. Pricing is per-token, failed pages are not charged, and the service costs $0.86 to $48.82 per 1,000 pages depending on the model.
Microsoft Research Asia has open-sourced Agent Lightning v1.0, a roughly 3,500-line agentic RL framework that trains the same agent harness used in deployment. In an end-to-end coding agent pipeline, Qwen3.5-9B rose from 41.8% to 56.4% Pass@1 on SWE-bench Verified using about 6,000 training samples. The framework runs agents as standard Kubernetes jobs without paid commercial sandbox services.
AIWhy it matters: The source shows how training with the deployed agent harness avoids rebuilding agents, and reports concrete SWE-bench Verified gains from about 6,000 samples.
We're making new Stage 1 checkpoints available with the byte-level components already trained & the original model's weights unchanged. Researchers can build on these to test new architectures & train the full system without repeating that initial stage.